public inbox for kvm@vger.kernel.org
 help / color / mirror / Atom feed
From: Alexandru Elisei <alexandru.elisei@arm.com>
To: pbonzini@redhat.com, thuth@redhat.com, andrew.jones@linux.dev,
	kvm@vger.kernel.org, kvmarm@lists.cs.columbia.edu,
	nikos.nikoleris@arm.com
Subject: Re: [kvm-unit-tests RFC PATCH 00/19] arm/arm64: Rework cache maintenance at boot
Date: Sun, 5 Nov 2023 10:16:50 +0000	[thread overview]
Message-ID: <ZUdrku2u6OGc43wv@monolith> (raw)
In-Reply-To: <20220809091558.14379-1-alexandru.elisei@arm.com>

Hi,

I had a v2 almost finished a few months ago (maybe a year), but I parked
the patches when we decided that the best thing to do was to have UEFI
support in first, then rework the cache maintenance - that way, the changes
can be tested on baremetal, which is much more unforgiving than KVM.

Unfortunately I've been busy with other things and I don't know when I'll
be able to come back to this. So I've pushed the work-in-progress here [1],
with what I hope is a helpful changelog since v1.

If my memory serves me right, even though it is marked as wip, it was
running just fine under KVM and I was waiting for UEFI to test it on
baremetal before posting.

I don't know when I'll have the time to get back to the series, so if
anyone is still interested, feel free to pick it up. If someone does pick
up the series, they can drop/rework patches as they see fit, and I'll try
to do my best to review whatever is posted on the list (but no promises).

[1] https://gitlab.arm.com/linux-arm/kvm-unit-tests-ae/-/tree/arm-arm64-rework-cache-maintenance-at-boot-v2-wip2

Thanks,
Alex

On Tue, Aug 09, 2022 at 10:15:39AM +0100, Alexandru Elisei wrote:
> I got the idea for this series as I was looking at the UEFI support series
> [1]. More specifically, I realized that the cache maintenance performed by
> asm_mmu_disable is insuficient. Patch #19 ("arm/arm64: Rework the cache
> maintenance in asm_mmu_disable") describes what is wrong with
> asm_mmu_disable. A detailed explanation of what cache maintenance is needed
> and why is needed can be found in patch #18 ("arm/arm64: Perform dcache
> maintenance at boot").
> 
> Then I realized that I couldn't fix only asm_mmu_disable, and leave the
> rest of kvm-unit-tests without the needed cache maintenance, so here it is,
> my attempt at adding the cache maintenace operations (from now on, CMOs)
> required by the architecture.
> 
> My approach is to try to enable the MMU and build the translation tables as
> soon as possible, to avoid as much of cache maintenance as possible. I
> didn't want to do it in the early assembly code, like Linux, because I like
> the fact that kvm-unit-tests keeps the assembly code to a minimum, and I
> wanted to preserve that. So I made the physical allocator simpler (patches
> #2-#6) so it can be used to create the translation tables immediately after
> the memory regions are populated.
> 
> After moving some code around, especially how the secondaries are brought
> online, the dcache maintenance is implemented in patch #18 ("arm/arm64:
> Perform dcache maintenance at boot").
> 
> The series is an RFC, and I open to suggestions about how to do things
> better; I'm happy to rework the entire series if a better approach is
> proposed.
> 
> Why is this needed? Nobody complained about test failing because of missing
> CMOs before, so why add them now? I see two reasons for the series:
> 
> 1. For architectural correctness. The emphasis has been so far on the test
> themselves to be architectural compliant, but I believe that the boot code
> should get the same treatment. kvm-unit-tests has started to be used in
> different ways than before, and I don't think that we should limit
> ourselves to running under one hypervisor, or running under a hypervisor at
> all. Which brings me to point number 2.
> 
> 2. If nothing else, this can serve as a showcase for the UEFI support
> series for the required cache maintenance. Although I hope that UEFI
> support will end up sharing at least some of the boot code with the
> non-UEFI boot path.
> 
> This is an RFC and has some rough edges, probably also bugs, but I believe
> the concept to be sound. If/when the series stabilizes, I'll probably split
> it into separate series (for example, the __ASSEMBLY__ define patch could
> probably be separate from the others). Tested by running all the arm and
> arm64 tests on a rockpro64 with qemu.
> 
> [1] https://lore.kernel.org/all/20220630100324.3153655-1-nikos.nikoleris@arm.com/
> 
> Alexandru Elisei (19):
>   Makefile: Define __ASSEMBLY__ for assembly files
>   lib/alloc_phys: Initialize align_min
>   lib/alloc_phys: Use phys_alloc_aligned_safe and rename it to
>     memalign_early
>   powerpc: Use the page allocator
>   lib/alloc_phys: Remove locking
>   lib/alloc_phys: Remove allocation accounting
>   arm/arm64: Mark the phys_end parameter as unused in setup_mmu()
>   arm/arm64: Use pgd_alloc() to allocate mmu_idmap
>   arm/arm64: Zero secondary CPUs' stack
>   arm/arm64: Enable the MMU early
>   arm/arm64: Map the UART when creating the translation tables
>   arm/arm64: assembler.h: Replace size with end address for
>     dcache_by_line_op
>   arm: page.h: Add missing libcflat.h include
>   arm/arm64: Add C functions for doing cache maintenance
>   lib/alloc_phys: Add callback to perform cache maintenance
>   arm/arm64: Allocate secondaries' stack using the page allocator
>   arm/arm64: Configure secondaries' stack before enabling the MMU
>   arm/arm64: Perform dcache maintenance at boot
>   arm/arm64: Rework the cache maintenance in asm_mmu_disable
> 
>  Makefile                   |   5 +-
>  arm/Makefile.arm           |   4 +-
>  arm/Makefile.arm64         |   4 +-
>  arm/Makefile.common        |   4 +-
>  arm/cstart.S               |  59 ++++++++++++------
>  arm/cstart64.S             |  56 +++++++++++++----
>  lib/alloc_phys.c           | 122 ++++++++++++-------------------------
>  lib/alloc_phys.h           |  13 +++-
>  lib/arm/asm/assembler.h    |  15 ++---
>  lib/arm/asm/cacheflush.h   |   1 +
>  lib/arm/asm/mmu-api.h      |   1 +
>  lib/arm/asm/mmu.h          |   6 --
>  lib/arm/asm/page.h         |   2 +
>  lib/arm/asm/pgtable.h      |  52 ++++++++++++++--
>  lib/arm/asm/thread_info.h  |   3 +-
>  lib/arm/cache.S            |  89 +++++++++++++++++++++++++++
>  lib/arm/io.c               |   5 ++
>  lib/arm/io.h               |   3 +
>  lib/arm/mmu.c              |  60 +++++++++++-------
>  lib/arm/processor.c        |   6 +-
>  lib/arm/setup.c            |  66 ++++++++++++++++----
>  lib/arm/smp.c              |   9 ++-
>  lib/arm64/asm/assembler.h  |  11 ++--
>  lib/arm64/asm/cacheflush.h |  32 ++++++++++
>  lib/arm64/asm/mmu.h        |   5 --
>  lib/arm64/asm/pgtable.h    |  67 ++++++++++++++++++--
>  lib/arm64/cache.S          |  85 ++++++++++++++++++++++++++
>  lib/arm64/processor.c      |   5 +-
>  lib/devicetree.c           |   2 +-
>  lib/powerpc/setup.c        |   8 +++
>  powerpc/Makefile.common    |   1 +
>  powerpc/cstart64.S         |   1 -
>  powerpc/spapr_hcall.c      |   5 +-
>  33 files changed, 608 insertions(+), 199 deletions(-)
>  create mode 100644 lib/arm/asm/cacheflush.h
>  create mode 100644 lib/arm/cache.S
>  create mode 100644 lib/arm64/asm/cacheflush.h
>  create mode 100644 lib/arm64/cache.S
> 
> -- 
> 2.37.1
> 

  parent reply	other threads:[~2023-11-05 10:16 UTC|newest]

Thread overview: 52+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-08-09  9:15 [kvm-unit-tests RFC PATCH 00/19] arm/arm64: Rework cache maintenance at boot Alexandru Elisei
2022-08-09  9:15 ` [kvm-unit-tests RFC PATCH 01/19] Makefile: Define __ASSEMBLY__ for assembly files Alexandru Elisei
2022-08-09 12:36   ` Nikos Nikoleris
2022-09-20  8:11   ` Andrew Jones
2022-08-09  9:15 ` [kvm-unit-tests RFC PATCH 02/19] lib/alloc_phys: Initialize align_min Alexandru Elisei
2022-09-20  8:20   ` Andrew Jones
2022-08-09  9:15 ` [kvm-unit-tests RFC PATCH 03/19] lib/alloc_phys: Use phys_alloc_aligned_safe and rename it to memalign_early Alexandru Elisei
2022-09-20  8:27   ` Andrew Jones
2022-08-09  9:15 ` [kvm-unit-tests RFC PATCH 04/19] powerpc: Use the page allocator Alexandru Elisei
2022-08-09  9:15 ` [kvm-unit-tests RFC PATCH 05/19] lib/alloc_phys: Remove locking Alexandru Elisei
2022-09-20  8:45   ` Andrew Jones
2022-09-20 13:20     ` Alexandru Elisei
2022-09-20 14:59       ` Andrew Jones
2022-09-26 15:04         ` Alexandru Elisei
2022-08-09  9:15 ` [kvm-unit-tests RFC PATCH 06/19] lib/alloc_phys: Remove allocation accounting Alexandru Elisei
2022-09-20  8:40   ` Andrew Jones
2022-09-20 13:19     ` Alexandru Elisei
2022-08-09  9:15 ` [kvm-unit-tests RFC PATCH 07/19] arm/arm64: Mark the phys_end parameter as unused in setup_mmu() Alexandru Elisei
2022-09-20  8:58   ` Andrew Jones
2022-09-26 11:01     ` Alexandru Elisei
2022-08-09  9:15 ` [kvm-unit-tests RFC PATCH 08/19] arm/arm64: Use pgd_alloc() to allocate mmu_idmap Alexandru Elisei
2022-09-20  9:05   ` Andrew Jones
2022-08-09  9:15 ` [kvm-unit-tests RFC PATCH 09/19] arm/arm64: Zero secondary CPUs' stack Alexandru Elisei
2022-08-09 12:56   ` Nikos Nikoleris
2022-08-10  9:42     ` Alexandru Elisei
2022-08-10 10:00       ` Nikos Nikoleris
2022-09-20  9:24   ` Andrew Jones
2022-08-09  9:15 ` [kvm-unit-tests RFC PATCH 10/19] arm/arm64: Enable the MMU early Alexandru Elisei
2022-08-09  9:15 ` [kvm-unit-tests RFC PATCH 11/19] arm/arm64: Map the UART when creating the translation tables Alexandru Elisei
2022-08-09  9:15 ` [kvm-unit-tests RFC PATCH 12/19] arm/arm64: assembler.h: Replace size with end address for dcache_by_line_op Alexandru Elisei
2022-08-09 13:01   ` Nikos Nikoleris
2022-09-20  9:37   ` Andrew Jones
2022-08-09  9:15 ` [kvm-unit-tests RFC PATCH 13/19] arm: page.h: Add missing libcflat.h include Alexandru Elisei
2022-09-20  9:39   ` Andrew Jones
2022-09-26 11:02     ` Alexandru Elisei
2022-08-09  9:15 ` [kvm-unit-tests RFC PATCH 14/19] arm/arm64: Add C functions for doing cache maintenance Alexandru Elisei
2022-08-09  9:15 ` [kvm-unit-tests RFC PATCH 15/19] lib/alloc_phys: Add callback to perform " Alexandru Elisei
2022-08-09  9:15 ` [kvm-unit-tests RFC PATCH 16/19] arm/arm64: Allocate secondaries' stack using the page allocator Alexandru Elisei
2022-09-20  9:58   ` Andrew Jones
2022-08-09  9:15 ` [kvm-unit-tests RFC PATCH 17/19] arm/arm64: Configure secondaries' stack before enabling the MMU Alexandru Elisei
2022-08-09  9:15 ` [kvm-unit-tests RFC PATCH 18/19] arm/arm64: Perform dcache maintenance at boot Alexandru Elisei
2022-08-09  9:15 ` [kvm-unit-tests RFC PATCH 19/19] arm/arm64: Rework the cache maintenance in asm_mmu_disable Alexandru Elisei
2022-08-09 13:53   ` Nikos Nikoleris
2022-08-09 14:22     ` Alexandru Elisei
2022-08-09 15:53       ` Nikos Nikoleris
2022-08-09 16:53         ` Alexandru Elisei
2022-08-09 19:48           ` Nikos Nikoleris
2022-08-10  8:52             ` Alexandru Elisei
2022-08-09  9:49 ` [kvm-unit-tests RFC PATCH 00/19] arm/arm64: Rework cache maintenance at boot Alexandru Elisei
2023-11-05 10:16 ` Alexandru Elisei [this message]
2023-11-06  9:37   ` Shaoqin Huang
2023-11-07  9:01     ` Alexandru Elisei

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=ZUdrku2u6OGc43wv@monolith \
    --to=alexandru.elisei@arm.com \
    --cc=andrew.jones@linux.dev \
    --cc=kvm@vger.kernel.org \
    --cc=kvmarm@lists.cs.columbia.edu \
    --cc=nikos.nikoleris@arm.com \
    --cc=pbonzini@redhat.com \
    --cc=thuth@redhat.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox